Custom Source

Introduction

Fumadocs is very flexible. You can integrate with any content source, even without an official adapter.

Examples

You can see examples to use Fumadocs with a CMS, which allows a nice experience on publishing content, and real-time update without re-building the app.

Building Content Source

For a custom content source implementation, there's two ways:

Using Loader API

Loader API has a file-system-like interface for integrating different content sources.

See Loader API source for details.

Using Low-Level API

For more robust control, you can use lower level APIs directly.

Page Tree

You can either hardcode the page tree, or write some code to generate one. See Definitions of Page Tree.

Pass your page tree to DocsLayout (usually in a layout.tsx):

layout.tsx

import { DocsLayout } from 'fumadocs-ui/layouts/docs';
import type { ReactNode } from 'react';

export default function Layout({ children }: { children: ReactNode }) {
  return (
    <DocsLayout
      tree={
        {
          // your own tree
        }
      }
    >
      {children}
    </DocsLayout>
  );
}

The page tree is like a smarter "sidebar items", they will be referenced everywhere in the UI for navigation elements, such as the page footer.

Page Content

For docs page, it's the same logic:

Define path params (slugs).
Fetch page content from path params.
Render the content (inside the DocsPage component).

You may need table of contents, which can be generated on your own, or using the getTableOfContents utility (Markdown/MDX only).

import { DocsPage, DocsBody } from 'fumadocs-ui/page';
import { getPage } from './my-content-source';
import { notFound } from 'next/navigation';

export default function Page({ params }: { params: { slug?: string[] } }) {
  const page = getPage(params.slug);
  if (!page) notFound();

  return (
    <DocsPage toc={page.tableOfContents}>
      <DocsBody>{page.render()}</DocsBody>
    </DocsPage>
  );
}

Pre-rendering

The mechanism for pre-rendering differs depending on your React.js framework.

It's important to pre-render pages (especially the frequently visited ones) to improve initial load time.

In Next.js

Define the generateStaticParams function to populate dynamic & catch-all routes.

When pre-rendering pages from a remote source, make sure to optimize repeated reads during the build phase. For example, if you're fetching content via GitHub API, it is better to clone the repository content locally for pre-rendering.

/docs/* -> GET github.com
/docs/introduction -> GET github.com
/docs/my-page -> GET github.com
Error: API Ratelimit

Document Search

This can be difficult considering your content may not be necessarily Markdown/MDX. For Markdown and MDX, the built-in Search API + search index usage is adequate for most use cases. Otherwise, you will have to bring your own implementation.

We recommend 3rd party solutions like Orama or Algolia Search. They are more flexible than the built-in Search API, and is easier to integrate with remote sources. Fumadocs offers a variant of Search Integrations.

MDX Remote

Fumadocs offers the MDX Remote package, it is a helper to integrate Markdown-based content sources with Fumadocs. You can think it as a next-mdx-remote with built-in plugins for Fumadocs.

Setup

npm install @fumadocs/mdx-remote

The main feature it offers is the MDX Compiler, it can compile MDX content to JSX nodes. Since it doesn't use a bundler, there's some limitations:

No imports and exports in MDX files.

It's compatible with Server Components. For example:

import { createCompiler } from '@fumadocs/mdx-remote';
import { getPage } from './my-content-source';
import { DocsBody, DocsPage } from 'fumadocs-ui/page';
import { getMDXComponents } from '@/mdx-components';

const compiler = createCompiler({
  // options
});

export default async function Page({
  params,
}: {
  params: { slug?: string[] };
}) {
  const page = getPage(params.slug);
  const compiled = await compiler.compile({
    source: page.content,
  });

  const MdxContent = compiled.body;

  return (
    <DocsPage toc={compiled.toc}>
      <DocsBody>
        <MdxContent components={getMDXComponents()} />
      </DocsBody>
    </DocsPage>
  );
}

Images

On serverless platforms like Vercel, the original public folder (including static assets like images) will be removed after production build. The MDX compiler might no longer be able to access local images in public.

When referencing images, make sure to use a URL.

On this page